Abstract
Scatter plots of satellite multispectral microwave emissivities at two or three different frequencies and/or polarizations reveal unique clusters that are persistent throughout winter conditions in the Arctic region. Cluster analysis of both global and regional data shows that there are more than one radiometrically different types of multiyear ice in the central Arctic, reflecting the strong effect of volume scattering, which varies with the size and shape of the inhomogeneities in the material as well as with the overall structure and history of formation. Some of the clusters in the central Arctic also exhibit “hook patterns” in 18‐GHz versus 37‐GHz plots, indicating sensitivity to modification of the signature by some surface and subsurface effects at 37 GHz. Sensitivity to snow cover and other surface effects in the first‐year ice region is also evident at 37 GHz. A coplanarity test using three SMMR channels shows the potential as well as limitation of the sensor in discriminating different ice types. However, the capability for accurate determination of sea ice concentration, especially in the central Arctic, is evident from scatter plots of the emissivities using the two polarizations at 37 GHz. The consolidated ice data in the central Arctic form a very compact and linear set of points in these plots, and they slope the same way throughout the winter of 1979 and in other years (1980–1984). An ice concentration algorithm based on the utilization of this slope (∼1.03) is developed that yields retrieved values with an accuracy of about 5% in the central Arctic and 10% in the seasonal sea ice region. The total areal ice extent and actual ice cover were also derived from ice concentrations retrieved by using the algorithm, and the results from several years of midwinter data show a relatively stable ice cover in the northern hemisphere.